[triclops200]

He was a super genius. Just a super genius who repeatedly called for genocide and nuclear first strikes. He was removed from the target selection committee for the Japan bombings because he kept advocating for high civilian targets with no military value and even the generals at the table viewed him as crazy. The ability to be a genius and a lunatic aren't mutually exclusive and the recent efforts to build a hagiography around him are concerning.

[vishho]

A first strike is not lunacy. It is a rational decision, if the nuclear arms race is viewed under game theory.

He was a consultant, not had his hand on the button. The people with the hand on the button were talking about the rising threat and how to deal with it in the future. Neumann rightly reasoned that if there ever came a conflict, the winning move would have been an early first strike.

If you look at his statements, these are statements of mathematical fact, not political strategy.

If anything, he showed the hypocracy of target selection and its justifications for war. You want to play to win, or worry about rules and perceptions, unsure of if your opponent employs similarly intelligent analysts and game theorists?

Look at the target selection of American adversaries in their information warfare. Made by lunatics? Or made to win a shadow war and damage American culture and politics without regard to military status?

[gonehome]

In the prisoner's dilemma isn't a core constraint that the prisoners can't collaborate? Therefore you can't cooperate for the best outcome and both are independently incentivized to defect.

In global diplomacy the goal is to coordinate for the best outcome. So I'm not sure a strike first 'defection' is the rational move.

[vishho]

Game Theory eventually settled on Mutually Assured Destruction. MAD can be seen as the tit-for-tat strategy in an iterated PD with multiple level thinking (If we have crazy von Neumann, do they have similar consultants? Do they know that we think Neumann is a lunatic? etc.) and super rationality (we are all players on the same board, damaging other players reduces our own chances to grow, we should be competing against those urging for war, on either side).

Von Neumann was consultant on game theory, not on geopolitical diplomatic strategy. War generals wanted to talk about future conflicts. Von Neumann reminded them that all talk of winning future conflicts could be made moot by a single move. And he reasoned that intelligent analysts on the other side would inform their generals similarly. Both did a good job, and the generals are commended for taking things outside pure maths into consideration.

What is our solution for when Russia joins the game and gets access to its own nukes?

The Spockian rational answer is: make this question irrelevant and increase our power on the board, by making sure Russia cannot even join the game we are currently winning.

To make good decisions you need diverse expert input like this. All in all, I think von Neumann's work helped keep the nuclear war on paper, instead of reality. His input of a first strike evolved into MAD and allowed us to reach an equilibrium.

[beerandt]

Also, for any of this to work, no matter the theory or rationale or lack thereof, the otherside has to believe that you're both capable and willing to execute any given plan.

Meaning that floating crazy-but-rationlized war strategies, especially sourced from individuals who are respected and known to be influential, can itself be seen as part of the overall strategy.

Regardless of whether said strategies are officially adopted or not. It's posturing.

[Reuzel]

Exactly. Trump weaponized this principle, making very rash and sudden moves during negotiations. This disadvantaged others, because they had trouble predicting his reaction to their own actions, thus moderating them in his favor or "You're fired!". I think a US general took it upon himself to inform China they would not be attacked, no matter what Trump threatened, because he himself was unable to model Trump's mind. "My task at the time was to de-escalate".

Also why MAD does not work well against information warfare. Is the current polarization of culture and politics a natural outgrowth of American culture, the result of unwitting civilians being targeted by military black and grey propaganda, or an unentangleable combination of the two? Did the opponent push the button? Did we push it back in the 80s and did they notice? Where exactly do we stand and draw the line, allowing countries to defend their (cultural) borders and feel safe, without the constant threat and fallout from offenders, who act like children pushing their parents to see how far they can go.

Sometimes I suspect these larger than life scientists working on the top-secret projects, Turing, Feynman, Shannon, Neumann, Kolmogorov, Tesla, Satoshi, were actually collections of people working undercover, an Alan Smithee catch-all type to launder intelligence, take credit, while keeping it in the shadows. Like the unnamed people supporting Bobby Fisher in his match against Russia. Modern day equivalents would be companies like Google, Dell, IBM, and Microsoft.

The moment the bomb became a button, was the moment the physicists had to step aside, and let the decision theorists step in. The bomb effectively became about how to make winning decisions. Decision science itself weaponized, opponents worrying about the analysts on the other side, not the aviators: you would know they would follow orders, and drop the bomb if instructed. You never fully know what those instructions were going to be, but you wanted to find out. Strategic Cold war espionage and misinformation must have ran wild.

[dragontamer]

There's no such restriction in the prisoner's dilemma.

Lets make it explicit. Let me _pretend_ that I'm going to work with you on the prisoner's dilemma game. I pretend that I'm trustworthy, and coordinate with you that we both trust each other.

That's when I betray you and take all the money for myself.

------------

In an *iterated* prisoner's dilemma, I simply work with you for the first X-1 trials (where X is the number of trials), and then betray you on the final trial. Since you know I'm going to betray you on the final trial, you betray me on the 2nd to last final trial. Etc. etc. This follows like induction all the way to the 1st trial.

Which means my best move is to betray you in the 1st trial, as per the rule of induction. But you know, pretend that I'm going to work with you (so that you choose trust in the 1st trial).

---------

Where things go sideways is iterated + public prisoner's dilemma. That is, there are 3 players, and the 3rd player watches what the other 2 players do. Each iteration, we rotate who plays the game.

Finally, we have a situation where trust is a good move: if only to "prove" to the other player that you're trustworthy and possibly get more gains over the long term.

[dilyevsky]

He was a living proof that general ai is extremely dangerous as he was basically Skynet from Terminator movies in human form.